Market inefficiency in person-to-person betting: examining 'known loser' insider trading on the exchanges.

• Author: Marginson, David
• Position: Symposium

1. Introduction

   Empirical tests of the efficient market hypothesis indicate that bookmaker-dominated horserace betting markets are weakly efficient at best (for example, Hausch, Ziemba, and Rubinstein 1981; Asch, Malkeil, and Quandt 1984; Crafts 1985; Gabriel and Marsden 1990). (1) Efficiency is undermined by 'market biases,' particularly the 'favorite-longshot bias' (Shin 1991, 1992, 1993), (2) which is commonly attributed to (racetrack) bookmakers' attempts to insure against insider trading (Schnytzer and Shilony 1995: Law and Peel 2002). Smith, Paton, and Vaughan Williams (2006, p. 1) applied the Shin model of market efficiency to the relatively new person-to-person internet betting exchanges (hereafter "exchanges") (3) and found "significantly lower market biases." The results are interpreted as evidence that insider trading on the exchanges "is not widespread" and "not as commonplace ... as is sometimes portrayed in the media" (Smith, Paton, and Vaughan Williams 2006, pp. 12-3).

   This study has two objectives. I first aim to show, through reference to exchange betting, the limitations of using the popular Shin (1991, 1993) measure of racetrack efficiency for detecting and measuring the incidence of insider trading in markets where traditional notions of bookmaking do not apply. A potentially more appropriate measure is presented in this context. The second and primary objective is to test for evidence of exchange-based insider trading, specifically insider activity aimed at profiting not from expected winners but from 'known losers.' This includes profiting from horses that are unplaced.

   Insiders can profit from knowing which horses in any given race are going to lose, (4) because (i) it is far easier to stop a horse from winning or from being placed than it is to ensure that it wins (Crafts 1985), and (ii) the exchanges provide unprecedented opportunities for unlicensed individuals to bet on horses to lose (O'Connor 2007a). (5) These two "facts" combine to generate unprecedented incentives for fraudulent practices (the laying of known losers), which are to some degree testable empirically using ideas presented by Crafts (1985). (6) The empirical test is conducted on a sample of 994 U.K. horse races that took place during 2005 and 2006. Results indicate that insider activity aimed at profiting from "known losers" may be commonplace on the exchanges. This study offers new insight into the efficiency of betting markets. The findings hold non-trivial regulatory implications.

   The article proceeds as follows. Section 2 considers the notion of insider trading and its analysis in the context of horserace betting. Section 3 outlines the characteristics of person-to-person internet betting markets. Section 4 summarizes the concerns raised in regard to betting media, which permit non-licensed individuals to 'bet on horses to lose.' Section 5 explains the approach used to detect and measure known loser insider activity on the exchanges. Section 6 describes the study's methodology and analysis. Section 7 presents the empirical results, and section 8 concludes.

2. Insider Trading

   Insider trading is difficult to define, but may generally be viewed as an activity by which those with access to non-public information seek to profit from this 'inside' information (Dowie 1976). Insider trading has a generally bad reputation (see, for example, Dowie 1976; Jarrell and Poulsen 1989; Shin 1993) and is proscribed as fraudulent in many capitalist countries. (7) Despite this, insider trading is considered an inevitable feature of horserace betting (see, for example, Figlewski 1979; Crafts 1985; Shin 1993: Law and Peel 2002; O'Connor 2007b); although, studies have yet to examine insider activity aimed at profiting from known losers.

   Attempts to measure the extent of insider trading within the bookmaker-dominated betting market have involved two different approaches. One method (discussed further in section 5) is to assess insider activity by reference to "significant market movers," often referred to as "plungers" or "steamers" (Crafts 1985; Schnytzer and Shilony 1995; Law and Peel 2002). The other and more favored methodology has been to use market inefficiencies as a proxy for detecting and measuring the incidence of insider trading (for example, Vaughan Williams and Paton 1997; Smith, Paton, and Vaughan Williams 2006). For instance, the Shin measure of market efficiency assumes that the favorite-longshot bias results from monopoly layers' (bookmakers') response to the presence of insider traders (Shin 1991, 1993). Specifically, bookmakers adjust their price-setting procedure by reducing the odds offered on longshots (low probability, high returns) relative to the odds offered on favorites (high probability, low returns) (Schnytzer and Shilony 1995). (8) They may do so in order to pass the costs arising from insider activity on to outsiders; that is, recreational bettors who, according to popular explanations of the favorite-longshot bias, over-bet longshots and under-bet on favorites (Smith, Paton, and Vaughan Williams 2006) because they are risk loving (Rosett 1965) and/or because transaction and information costs restrict bettors' ability to calculate horses' "true" winning probabilities (Hurley and McDonough 1995). (9)

   The favorite-longshot bias is empirically well established (see, for example, Henery 1985; Thaler and Ziemba 1988; Vaughan Williams and Paton 1997; Bruce and Johnson 2000b), and the phenomenon continues to attract the attention of researchers (see, for example, Gramm, McKinney, and Owens 2008; Winter and Kukuk 2008). In a recent study, Smith, Paton, and Vaughan Williams (2006) apply the Shin z measure to betting exchange data and report a significantly lower favorite-longshot bias of 0.09%, compared to a racetrack bias of 2.17%, for their 700-race sample. Based on Shin's analysis (Shin 1991, 1992, 1993), these results indicate that betting exchanges may have brought about significant efficiency gains compared to bookmaker-dominated horserace betting markets, with the implication that the exchanges are not problematized by insider trading, or at least are problematized less so than racetrack betting. At the same time, studies that tie the examination of insider trading to Shin's z measure make the assumption that a reduction in (or absence of) market inefficiencies, such as that seen with the favorite-longshot bias, signifies a reduction in (or absence of) insider activity. This assumption may possibly hold for bookmaker-dominated markets. However, in outlining the characteristics of exchange betting, the next section brings into question the appropriateness of using the Shin z measure of insider trading to assess the efficiency of betting markets where traditional notions of bookmaking do not apply.

3. Exchange Betting

   Traditional notions of bookmaking do not apply on the exchanges because betting here is "betting without bookies" (O'Connor 2007b). Licensed bookmakers fulfill a crucial market maker role in traditional betting markets; their license permits them, and only them, to lay horses to lose. Non-licensed individuals are restricted to betting on horses to win. There is no such restriction on the exchanges. As Smith, Paton, and Vaughan Williams (2006, p. 2) explain, "Betting exchanges exist to match people who want to bet on a future outcome at a given price with those who are willing to offer that price. The person who bets on the event happening at a given price is the 'backer' [or 'bettor']; the person who offers the price is known as the 'layer.'"

   For example, someone may wish to lay a particular horse at, say, odds of 3 to 1 to a maximum stake of, say, 100 [pounds sterling] (maximum liability is 300 [pounds sterling]). Someone else, the bettor, may accept these odds and place a bet with the layer, say 100 [pounds sterling] at 3 to 1. A wager thus occurs between the two parties. Equally, exchange bettors can indicate (i) the odds at which they would be willing to bet on a particular horse (say 7 to 2) and (ii) the magnitude of the wager at that odds level (say 100 [pounds sterling]). Anyone wishing to lay the same horse at those odds can accept this bet and form a transaction with the bettor. A wager thus occurs once more--between layer and bettor.

   In essence, betting exchanges may be viewed as comprising multiple two-person contracts. No market maker (bookmaker) is involved, and importantly, no one is obliged to "maintain a credible market structure embracing all runners" (Smith, Paton, and Vaughan Williams 2006, p. 14). As such, the notion of an engineered favorite-longshot bias has no meaning on the exchanges. Rather, the layer on the exchanges can only attempt to manipulate the odds offered on the individual horse(s) in any given race. (10) Given this, and assuming bookmakers are responsible for engineering the favorite-longshot bias, (11) it follows that it may be inappropriate to use the Shin measure (which presupposes the presence of licensed bookmakers) to test for insider trading on the exchanges, which are characterized by the absence of bookmakers. Basically, the Shin measure may be insufficiently sensitive to the characteristics of exchange betting, leading to possible misinterpretation of the results obtained (Bruce and Johnson 2000b).

   This point is developed in section 5, which aims to show, as part of the analysis presented, that the type of insider activity facilitated by the exchanges--profiting from known losers-may actually create a paradox if the Shin measure is used to assess the efficiency of the exchanges: Known loser insider activity may give rise to the illusion of increasing efficiency in the context of increasing inefficiency. The import of evaluating the efficiency of the exchanges is highlighted in the next section, which outlines concerns raised about the unprecedented opportunities internet person-to-person betting provides for fraudulent practices in the...